IMES DISCUSSION PAPER SERIES INSTITUTE FOR MONETARY AND ECONOMIC STUDIES BANK OF JAPAN C.P.O BOX 203 TOKYO 100-8630 JAPAN What is Systemic Risk? Moral hazard, initial shocks and propagation James Dow Discussion Paper No. 2000-E-17
IMES DISCUSSION PAPER SERIES
INSTITUTE FOR MONETARY AND ECONOMIC STUDIES
BANK OF JAPAN
C.P.O BOX 203 TOKYO
100-8630 JAPAN
What is Systemic Risk?
Moral hazard, initial shocks and propagation
James Dow
Discussion Paper No. 2000-E-17
NOTE: IMES Discussion Paper Series is circulated in
order to stimulate discussion and comments. Views
expressed in Discussion Paper Series are those of
authors and do not necessarily reflect those of
the Bank of Japan or the Institute for Monetary
and Economic Studies.
IMES Discussion Paper Series 2000-E-17June 2000
What is Systemic Risk?Moral hazard, initial shocks and propagation
James Dow*
Abstract
This paper discusses different aspects of the notion of systemic risk. It contains aselective survey of the research on related topics, a review of some case studies offinancial crises and failures, and a discussion pointing towards the importance ofmoral hazard as a key element of systemic risk. The main ideas studied are thelinks between: capital structure theory and bank capital regulation; moral hazardand agency theory at the level of the individual trader, the financial firm, and theoverall financial system. Another important idea is the co-determination of assetprices and bank solvency. My main focus is on moral hazard as a potentiallyfruitful area for future research. Although existing research emphasizes thepowerful propagation mechanisms whereby a small initial shock can be amplifiedby the financial system, I suggest that moral hazard, together with leverage at thelevel of the individual firm, can cause a large shock to the financial system.
* London Business School (E-mail: [email protected] / Home Page: http://www.lbs.ac.uk/faculty/jdow)
This paper was prepared for the Bank of Japan. I thank Urs Birchler of the Swiss National Bank,
Xavier Freixas of Universitat Pompeu Fabra, Shuji Kobayakawa of the Bank of Japan, Stephen
Schaefer of London Business School, Oren Sussman of Ben Gurion University, and an anonymous
referee, for helpful comments on an earlier draft. Any opinions expressed in this paper are my own
and do not represent the views of any of the institutions or people referred to here.
Contents
1. Introduction
2. Different notions of systemic risk
3. Review of selected research literature
4. Central themes in research on systemic risk
5. Review of selected cases
6. Moral hazard
7. Relative importance of different kinds of systemic risk
8. Conclusions
9. References
1
1. Introduction
The purpose of this paper is to discuss different notions of systemic risk, and to clarify the main
ideas and policy issues relating to systemic risk. Although it is partly a survey, it also explores
ideas for new directions in research on systemic risk. It is idiosyncratic rather than
comprehensive in the sense that, although I cover quite a range of topics (existing research
literature, case studies, and ideas on systemic risk), the selection of topics reflects a personal
view on what is relevant.
The problem of systemic risk is of great interest to public policy makers, notably central bankers
since the prevention of systemic risks is one of the main tasks of the central bank. The 1998
financial crisis in developing countries in Asia, Russia and Latin America underlines that this
concern is justified. Although the more developed economies have largely escaped damage in
this crisis, concern about the present high level of equity markets, notably in the USA, reflects
the potential for problems there as well.
Despite this, there is not a fixed, universally recognized definition of systemic risk. The phrase
is used to describe several different kinds of financial crisis. For example, it can be used in a
narrow sense to describe problems in the payments system, or at the other extreme in a very
broad sense to describe a financially-driven macroeconomic crisis. In this paper I shall attempt
to address this deficiency, not by providing a comprehensive analysis but by discussing the
meaning of systemic risk and by discussing how existing economic models can be related to the
problem of systemic risk.
2
The organization of the paper is as follows. I start by discussing several different notions of
systemic risk (section 2). My starting point is the idea that systemic risk, being a matter of
public policy, should refer to cases of risks being imposed on the financial system where some
element of externality exists. In other words, financial regulators have a legitimate interest in
intervening when somebody takes a risk that then causes a further risk for others in the financial
system. This externality means that overall risk-taking may be excessive.
I next survey some selected research that bears on the topic (section 3) and synthesize and
discuss the main ideas that emerge from this literature (section 4).
Having explored the academic literature on systemic risk, I look at the problem in another way
by turning to a small number of cases of financial crises of regulatory interest, including both
cases where an individual firm was the source of the problem, and cases of more general failure
of the financial sector (section 5). Some of these are well-known cases, others less so. These
cases are not selected to have any special common features, nor to support any particular theory.
From them I infer that moral hazard, broadly defined, is an important element of financial crises.
This motivates a discussion of problems of moral hazard in relation to financial failure. I
discuss different kinds of moral hazard problem in section 6, both at the individual level and at
the collective level. I then return (section 7) to a more comprehensive review of different cases
of financial failure, discussing which aspects of moral hazard are of greatest importance in these
cases. My main conclusion is that straight moral hazard of the kind usually studied in
economics is less important than a kind of collective moral hazard in which the corporate
culture is flawed in terms of incentives, either by being too weak and passive or by encouraging
3
aggressive individual behaviour that may not be in the collective interest. Section 8 concludes.
The main departure in this paper from the existing literature on systemic risk is that I emphasize
the origin of the shocks that can cause financial crises, rather than the propagation mechanisms
whereby a small shock to one firm can spread to other firms and be amplified into a large shock
to the whole system. Implicitly, the approach suggested here is that initial shocks can be
multiplied inside the firm by moral hazard and financial leverage, so that an individual firm may
deliver a large shock to the financial system. Additionally it may be that many firms take
similar actions and hence may deliver highly correlated shocks. The LTCM crises is an
example. I do not suggest that the propagation mechanisms are unimportant, but that it may
also be fruitful to consider the initial shocks that may trigger systemic risk.
4
2. Different Notions of Systemic Risk
The term "systemic risk" has been used to cover different types of economic and regulatory
problems. I shall discuss four of these different meanings, to provide an organizing framework
for the ideas developed later in this paper.
Systemic Risk and Depositor Protection
Research in banking has focused heavily on the role of deposit insurance in influencing the
development of the banking industry (for a survey, see Freixas and Rochet, 1997). This deposit
insurance may be explicit, or it may be implicit -- banks that are too big to fail provide an
example of institutions whose implicit insurance from the government is probably more
valuable to depositors than whatever explicit deposit insurance scheme may be in place.
Starting from Bagehot (1873), deposit insurance and the central bank's role as lender of last
resort have been identified as important instruments for preventing general bank panics as well
as individual bank runs. On the negative side, deposit insurance encourages asset substitution:
banks have an incentive to take excessively risky positions so that in the event of insolvency, the
losses will be borne by the deposit insurer rather than by either the shareholders or the
depositors. The Savings and Loan crisis in the US in the 1980's exemplifies these problems and
suggests a view of deposit insurance as an exogenous, politically-inspired distorting subsidy.
A different perspective is given in Dewatripont and Tirole's book (1994). They take the view
that some form of depositor protection is an important and desirable part of banking regulation.
Essentially their argument is that banks are just like other firms except that they are highly
5
leveraged with many small debtholders (i.e. the depositors). Since banks are highly leveraged,
their debtholders should monitor the management, but since the depositors are small and
dispersed it is impractical for them to do so. Hence the need for a regulatory agent (e.g. the
deposit insurer) to act in the depositors' interests.
Although depositor protection has been a dominant theme of academic research on banking
regulation, I shall not say much about it in this paper. There are two reasons for this. The first
is that I believe the existing research does not leave much unsaid. For example the standard
argument against depositor insurance as an incentive to asset substitution is clear and has been
made many times in the literature. For a comprehensive account of banking research, see
Freixas and Rochet (1997).
The second reason for concentrating on other aspects of systemic risk is that the adverse
incentive effects of deposit insurance can be reduced substantially by prompt closure of banks
that are approaching insolvency. This can occur in either of two ways. First, if regulators can
continuously monitor the capital adequacy of a bank they can intervene to force it to close
before its equity value falls to zero. Secondly, if the market can observe the same information
the bank will also be forced to close since it will cease to be able to find trading counterparties.
Improvements in transparency, caused by both regulatory measures such as the Basle
Committee capital adequacy rules and by private-sector developments such as better risk-
management systems, will surely make these mechanisms increasingly effective in the next few
years. For example, when Yamaichi securities closed in 1997, it was claimed to be marginally
solvent (in fact, it was probably insolvent, but still not by much when one considers that it
practised book-keeping irregularities to create hidden trading losses.) Although press reports
6
described the Yamaichi failure in predominantly negative terms, it seems on the contrary to
have been a relatively well-controlled closure: the firm was closed before it had accumulated
very large losses, and the closure was partially anticipated by the market and reflected in
increasing funding costs for the firm. If greater transparency and risk-management ensure that
future closures of banks and securities firms are as smooth as Yamaichi's, many aspects of the
problem of systemic risk will be much reduced.
Systemic Risk and Payments Systems
The disruption to the payments system caused by the default of one or more banks is often
viewed as a form of systemic risk. The chaos caused by the Herstatt failure illustrates some of
these problems. Much of the recent debate has focused on difference between gross and net
settlement: gross settlement avoids most of the problems of systemic risk, and progress in
information technology has made it much easier to implement.
As this is a rather technical subject, I start by explaining the differences between the institutions
of gross and net settlement. With deferred net settlement (DNS), payment orders are
accumulated at the settlement system over a suitable interval (e.g. a day) and at the end of the
period, the net amounts are settled either on a bilateral basis between pairs of banks (e.g. CHIPS
in the US), or on an overall basis between each individual member bank and the central system
organizer. With real-time gross settlement (RTGS) all payments are immediately settled by
making a transfer from the sending bank's account at the system organizer (usually the central
bank) to the payee's account. In many countries both DNS and RTGS co-exist. For example
the Bank of Japan's BOJ-NET system has both a DNS portion and a (much smaller) RTGS
7
alternative, while the US has both the privately organized CHIPS (DNS) and the Federal
Reserve's Fedwire (RTGS). Within RTGS systems there is an important dichotomy: on the one
hand, some systems do not allow the participants ever to overdraw on their accounts with the
organizer (as in the Swiss National Bank's SIC), or else they only allow fully-collateralized
credit (as in the UK's private-sector CHAPS). On the other hand, Fedwire allows
uncollateralized overdrafts and thus, effectively the Fed's lender-of-last resort function becomes
entangled with its role as a payments system operator.
If we focus more specifically on the differences between RTGS and DNS payments systems in
terms of systemic risk (i.e. ignoring the separate question of lender-of-last resort policy) the
appropriate comparison should be between DNS and RTGS without overdrafts (or with fully-
collateralized overdrafts). The traditional discussions in the literature (see the useful survey by
Holthausen (1997)) mostly suggest that RTGS has the benefit of reducing systemic risk at the
cost of requiring banks to hold higher reserves in their accounts with the payment system
operator. The incentive for banks to use DNS is illustrated by Fedwire's declining payment
volume relative to CHIPS: from 57% of the total in 1980, to 48% in 1990 and 42% in 1994
(Summers (1996)). However, being required to support RTGS payments with a proportionate
non-interest bearing reserve deposit at the central bank cannot be regarded as a social cost. It is
pure distortion that could be eliminated by paying market interest rates on the deposits or by
allowing banks the alternative of depositing government securities as collateral. Hence, moves
towards RTGS, such as the EU's TARGET system for the Euro currency zone, seem desirable.
Since the arguments concerning payments systems are somewhat separate to the other issues
which are the main focus of this paper, and since there is quite a well-developed literature on
8
payments systems, I shall not discuss them further. The reader is referred to Rochet and Tirole
(1996) and Freixas and Parigi (1997) for models of risk in payments systems, and to Holthausen
(1997) for a survey of the literature in this area.
Systemic Risk and Endogenous Asset Values
The third notion of systemic risk that I shall consider in this paper is linked to the idea that asset
values are determined endogenously in equilibrium. In general the assets of one bank depend
on the actions and the profitability of one or more other banks. For example if bank A has
deposits with bank B and B becomes insolvent, this will damage A, perhaps to the point that A
becomes insolvent also. Or, if bank A has real estate and also holds real estate mortgages, and
other large banks in the same economy are also involved in the real estate market, if these banks
decide to sell their holdings or foreclose on more of their loans, this will depress property prices
and damage bank A.
This effect could operate in more attenuated ways, as well as more extreme ways. For an
example of the former, we can consider what happens if a bank's value falls and, either to
maintain capital adequacy ratios or as part of a policy of optimal capital structure, it responds by
selling assets. This can depress asset values and hence set off the same process in other banks.
This weaker form of the effect illustrates that it is not necessary for any firm to default for asset
prices to be depressed by this mechanism. A more extreme version of this effect would be
when one firm defaults, sells assets and this depresses asset values to the point where another
firm may also default. This chain reaction is similar to what is often called the "domino effect"
of default.
9
Macroeconomic Aspects of Systemic Risk
The final type of systemic risk that I wish to distinguish is essentially a macroeconomic variant
of the previous one. The effect I just described was microeconomic in nature, but one can see
that similar effects could also arise in a macroeconomic context. Falling asset values, leading to
reduced capital adequacy for banks and lower collateral values for borrowers, could reduce
national income. Fisher's (1936) "Debt Deflation" is a classic analysis of the macroeconomic
role of debt; below I shall survey more recent research that captures some of these effects.
10
3. Review of Selected Research Literature
In this section I review some of the main research contributions that are related to the idea of
systemic risk. The papers are, for the most part, simply described and summarized individually
in this section. Later, in section 5, I relate the papers to each other and draw out what I believe
are the main themes that emerge from this research in relation to systemic risk.
The papers surveyed are: Blum and Hellwig's (1995) paper on macroeconomic implications of
capital regulation, Hellwig's (1995) work on intermediation and risk allocation, Allen and Gale's
(1998) model of financial crises, empirical work by Bernanke (1983) and others and theoretical
work by Kiyotaki and Moore (1997a, 1997b) and by Suarez and Sussman (1997a, 1997b) on the
role of credit in business cycles, and Holmstrom and Tirole's (1997) model of bank lending
cutbacks in recession.
There are, of course, many other papers that touch on the theme of systemic risk. The papers
surveyed here are chosen because they suggest certain common themes that will be elaborated in
the discussion in section 5 below.
3.1 Blum and Hellwig (1995): macroeconomic effects of capital adequacy rules
Blum and Hellwig's paper was inspired by the widespread international efforts (e.g. by the Basle
Committee on Banking Supervision) towards the introduction of bank capital requirements.
Their idea is that minimum capital requirements impose an upper bound on the amount of bank
lending for a given capital base. A negative shock to the economy, by reducing bank profits and
11
hence bank capital, will lead to reduced lending in subsequent periods. Hence, the effects of a
negative shock to the economic system will be amplified by the rigid link between banks'
balance sheets and their capital base. They show this using a standard macroeconomic model
similar to IS-LM.
3.2 Hellwig (1995): Hidden aggregate risk exposure
Hellwig has written several papers on risk and the role of banks in the economy, e.g. Hellwig
(1991, 1994a, 1994b, 1995, 1998). One point that emerges in several of these is that risk
regulation (such as the Basle Committee capital adequacy rules) may be defective because it
focuses only on the individual bank and not on the banking system as a whole. For an extreme
example, consider a banking system of 500 individual banks, where the n'th bank (all the way
from n=1 to n=500) has liabilities of duration n months and holds assets of duration n+1
(Hellwig (1995)). While each may look as if it is not highly exposed to the term structure, the
interest rate exposure of the banking system is large because it has transformed a 1-month
liability into a 40-year asset. For a more realistic example driven by similar principles, if banks
trade a large amount of swaps among themselves, there will be a low "signal-to-noise" ratio as
interbank swap volume is large relative to fundamentals. Individual risk exposures will not
reveal the exposure of the whole system.
Similar effects could also arise through inadequate modelling of risk exposure in the economy.
For example, a bank might consider itself hedged against interest rate risk, but could still be hit
by an interest rate rise that causes debtor insolvencies. Thai banks might borrow in US dollars
and lend in dollars to domestic firms that subsequently turn out to have no dollar income.
12
3.3 Allen and Gale (1998): bubbles and crises
Allen and Gale attempt to model the role of asset price bubbles in financial crises. Their model
has two main ingredients. First, financial institutions engage in asset substitution thereby
bidding up the price of risky assets and causing a bubble. Secondly, credit expansion enhances
the effect of asset substitution both contemporaneously and in anticipation.
Allen and Gale present two versions of their model. In the simpler version, banks with a fixed
supply of funds to lend interact with firms or individuals that want to borrow. Borrowers can
choose either safe investments or risky ones and, because of the limited liability option, will
tend to prefer risky investments: in other words, asset substitution will occur. In their model the
supply of risky investments is fixed, hence the increased flow of cash into risky investments will
tend to drive up the price. Both the price of the risky asset and the incidence of default are
higher than they would be in the absence of asset substitution. The authors interpret the higher
asset prices as a bubble and the higher chance of default as a financial crisis.
In the more complex version of the model, the asset substitution problem becomes more severe
in a multi-period setting. Uncertainty about the supply of funds for investment (interpreted as
being under the control of government monetary policy) introduces an extra element of risk and
increases the value of the limited liability option for borrowers. This increases the size of the
bubble and the chance of crisis.
Unlike most of the other papers reviewed here, Allen and Gale do not model the spillover effect
of the financial sector on the real economy. Hence, although they implicitly take for granted
13
that financial crises will have real effects, this is not part of their analysis.
3.4 Bernanke et al.: the credit channel
Traditional (neo-classical) macroeconomic analysis has downplayed the role of economic
institutions in economic crises (this also applies to the more recent real business-cycle theory).
In particular, Friedman and Schwartz's (1963) study of the US financial crisis of 1930-33
suggested that the primary way in which bank failures made the recession worse was through
the "monetary channel," i.e. the banking crisis led to a rapid contraction in the supply of money
which in turn affected national income. Bernanke (1983) argues that, in addition to their role in
the monetary channel, banks have an important role in producing and processing information as
part of their business of making loans. Hence, a destruction of part of the banking system as a
result of bank failures has real economic costs just as the destruction of physical assets would
(note that one third of US banks failed over the period 1930-33). In addition to the effects of
bank failures, surviving banks in the great depression made fewer loans, as they switched to
more liquid assets rather than holding loan portfolios. A third effect is caused by the nature of
loan contracts: a defaulting borrower must enter a costly and protracted bankruptcy process.
Together these effects suggest a "credit channel" for the real effects of the financial crisis.
Bernanke gives empirical evidence to support the existence of the credit channel.
Bernanke and Gertler (1989) present a theoretical model that represents a stylized view of how
the credit channel might operate. They argue that a borrower with a strong balance sheet is
effectively fully-collateralized and can borrow without the lending bank incurring any
monitoring costs, while a borrower with a weak balance sheet needs more intensive bank
14
monitoring. These monitoring costs are shifted onto the borrower in equilibrium in the form of
a higher interest rate, and hence may reduce investment as projects that are positive-NPV at the
riskless rate become uneconomic at the higher rate. Thus shocks to balance sheets, including
distributional shocks, can initiate macroeconomic fluctuations.
Some empirical evidence in favour of the credit channel is presented by Kashyap, Stein and
Wilcox (1993). They show that an overall contraction in firms' financing is accompanied by a
switch in short-term external financing from bank loans to commercial paper. This switch away
from bank financing also seems to have real effects on investment. This suggests that non-
intermediated financing is an imperfect substitute for bank financing.
Subsequent work in this field has concentrating on refining the idea that banks in recession will
switch to holding more liquid assets. This idea is often described as a "credit crunch" or a
"flight to quality." Gertler and Gilchrist (1993, 1994) find that in recessions, large firms in the
US appear to be able to continue financing inventories as sales fall, while small firms cut back
sharply. Similarly, Oliner and Rudebusch (1993) find that small firms are quicker to cut back
investment. These results are confirmed by more recent work by Bernanke, Gertler and
Gilchrist (1996) using disaggregated data on individual manufacturing firms. They also note
that, while small firms make up around half of the manufacturing sector in the US (by sales),
they account for over three quarters of sales in wholesale and retail trade, services and
construction.
3.5 Kiyotaki and Moore: credit cycles
15
Kiyotaki and Moore (1997a) construct a model in which firms' borrowing is constrained by the
collateral they can provide in support. The collateral consists of holdings of suitable fixed
assets such as land. A small negative shock to productivity reduces firms' profits and hence
their cash position, and if their collateral constraint is binding they will be unable to borrow
more to maintain investment. Hence investment, including investment in assets, must be
reduced. This will cause asset prices to fall, exacerbating the collateral constraint.
This describes a static multiplier process by which a temporary shock to productivity may have
a magnified effect on asset prices and hence on lending and output. However, there is also a
stronger dynamic element to the multiplier. The temporary shock to productivity has a
permanent effect on the profitability of firms whose borrowing is collateral-constrained, since in
each period their investment is reduced and hence their profits next period are also reduced --
repeating the initial process. Thus their demand for collateralizable assets will be permanently
reduced and this permanent fall in asset demand will be reflected in a much larger current drop
in asset prices.
These ideas are developed further in Kiyotaki and Moore (1997b) and in Kiyotaki (1997). One
problem with the analysis in Kiyotaki and Moore (1997a) is that, since the model is highly
complex and hence, the initial shock is modelled as an unanticipated disturbance, rather than the
outcome of an random variable whose uncertainty is anticipated by firms. Another problem is
that the contracts are suboptimal in the sense that it would be better to index the repayment
promised to the level of asset prices. In current research, Kiyotaki and Moore are working a
fully-specified stochastic model that extends the analysis to address both of these problems.
16
3.6 Suarez and Sussman: endogenous financial cycles
The research described above (Bernanke et al., and Kiyotaki and Moore) describes how the
financial sector may have a significant role in initiating and propagating recession, and in
amplifying the effect of shocks to the economy. However, these accounts are incomplete in that
they do not provide an account of how an economy may pass in and out of successive recessions
over time. There is a propagation mechanism (the "credit channel") but no reversion
mechanism. Suarez and Sussman (1997a) address this by providing a model of a highly stylized
and simplified economy which cycles in and out of recession over time, and where the financial
sector is the main mechanism driving these cycles. They concentrate on describing a reversion
mechanism rather than the propagation mechanism.
The model is based on an extension of the Stiglitz and Weiss (1981) model of credit, extended
to two periods. The Stiglitz and Weiss model (which also served as a motivation for Bernanke's
work on the credit channel) is based on the idea that, from a bank's point of view, borrowers'
creditworthiness tends to fall as the interest rate demanded increases (this can happen either
because of adverse selection or because of moral hazard). This leads to inefficiencies in funding
investment by bank loans: for example credit rationing may occur in equilibrium. For example,
Suarez and Sussman suppose that firms are run by entrepreneurs whose effort level affects the
viability of the firm's projects. External financing reduces their incentives to provide effort.
Cycles in their model are cause by the following mechanism. A boom results in high output of
both capital and consumption goods, leading to lower prices. This in turn reduces firms'
profitability and worsens the incentive problems caused by external finance, causing increased
17
business failures and recession. As a result prices rise, restoring profitability to surviving firms,
reducing incentive problems and causing a boom again. They show that, even in the absence of
any external shocks, their economy will display cycles.
While the model is arguably oversimplified in many ways, two virtues are that it endogenizes
the price of capital goods as part of the equilibrium, and that it considers optimal contracting
arrangements for external financing rather than exogenously-specified contractual forms such as
a standard bank loan agreement.
These ideas are developed further in Suarez and Sussman (1997b), a related paper which places
greater emphasis on the role of capital goods prices in financially-intermediated business cycles.
The mechanism they suggest is as follows: firms that invest in a boom will buy capital goods at
a high price, hence they need more external finance and this debt burden will give them a higher
likelihood of bankruptcy. The resulting wave of liquidations will result in lower prices of
capital goods, and firms that invest then will have a lower chance of default, leading to a boom.
The paper also examines the effect of different insolvency codes on the economy in equilibrium,
concluding that, surprisingly perhaps, a more debtor-oriented code (such as Chapter 11
bankruptcy in the US) will accentuate the economy's cyclicality.
An interesting application of this analysis that suggests itself (although not mentioned by the
authors) would be to use it to study the real-estate market. Beyond that, the advantage of this
model is that, although more stylized than the other analyses reviewed here, it provides an
equilibrium description of the complete cycle and not just of the onset of recession.
18
3.7 Holmstrom and Tirole (1997): the balance-sheet channel
Like many of the papers described above, Holmstrom and Tirole (1997) is a model where banks
are intermediaries whose economic role is to reduce informational asymmetry. The main
difference is that it also describes the role of banks' balance sheets on the patterns of financing
in the economy. Hence it can separately represent a "balance sheet channel" and a "lending
channel" for the real effects of financial intermediation. In the "balance sheet channel," if the
value of firms' collateral is reduced it will be harder for them to raise external finance, and if
they do they will be forced to rely more heavily on bank finance rather than (cheaper) securities
issues. A contraction via the "lending channel" (a reduction in bank capital) reduces both the
accessibility of external finance and the relative use of intermediated finance vs. the issue of
debt securities.
As in much of the literature, a crucial initial assumption is the difficulty of raising equity finance
(for either banks or other firms). Although there are various theoretical models that can be
invoked to help justify this assumption (e.g. Myers and Majluf (1984)), this remains an open
and important question for research. Holmstrom and Tirole conclude their paper with a
discussion of this point.
19
4. Central Themes in Research on Systemic Risk
A number of central themes emerge from the above discussion of the research literature. Here I
shall discuss what seem to me to be the main ones.
4.1 Institutions Do Matter
Traditional (neo-classical) economics takes the view that an economy is determined by its
"fundamental" characteristics such as individual preferences, technological possibilities and
resource endowments. An institution such as a bank or a firm is merely a "veil" that covers, but
does not hide, the underlying form of the economy's production function. There are two
different instances of this point of view that are relevant to the above literature.
The first arises in macroeconomics and is exemplified by the traditional "monetary channel"
view of the role of the banking system in recessions. This view is that the main channel for
propagating the recession is the aggregate money supply (the Friedman and Schwartz view of
the US great depression only admits a role for institutions in so far as they consider the
degeneration of the clearing houses, caused by the Fed, to have led to the reduction in the
money supply). In contrast, the more recent research has emphasized that financial
intermediaries have a real economic function in producing information and in monitoring
borrowers. Note that the analytical tool used to model this function is the economics of
asymmetric information.
The second instance is in relation to the Modigliani-Miller theorem, which views the firm as a
20
transparent vehicle for the claim-holders to own the assets of the firm. Much of the research
described above (Bernanke (1983), Bernanke and Gertler (1989), Kiyotaki and Moore (1997),
Suarez and Sussman (1997a, 1997b), etc) is based on violations of this principle, using
asymmetric information as the reason why firms cannot costlessly raise external finance.
However, from the point of view of systemic risk it is also important that the M&M theorem
cannot apply to banks either: for example, discussion of capital adequacy regulations makes no
sense unless we acknowledge that banks find equity expensive relative to other forms of capital
(Schaefer (1990)). Holmstrom and Tirole (1997), Blum and Hellwig (1995) and Hellwig (1998)
explicitly start from the assumption that bank equity is limited. For a model of bank capital
structure, see Dow and Rossiensky (1997). For a discussion the applicability of the M&M
principle to banking, see Miller (1995).
Unfortunately, our understanding of why and when the M&M principles fail to apply is
incomplete. It is easy to see that a firm with a limited equity base may face a variety of
difficulties in raising debt; some ways in which this can happen are due to the presence of
asymmetric information. However, the reasons for limitations on equity finance are less
obvious (particularly if rights issues are used). Myers and Majluf (1984) suggest an explanation
based on divergent interests between existing and new shareholders. However, there are
theoretical problems with the foundations of this model (see Raposo (1998)).
4.2 Endogenous Asset and Collateral Values
Several papers reviewed above feature models where firms require collateral to raise finance,
and collateral values are endogenously determined in equilibrium. We can view the effect of
21
endogenous asset values at two levels. First, at the level of non-financial firms, their real asset
values are endogenously determined, as in Kiyotaki and Moore (1997) and Suarez and Sussman
(1997a, 1997b). Secondly, similar effects may operate at the level of banks. For example, if a
bank is capital constrained, or is closed, while holding a large portfolio of property mortgages or
business loans the result might be a more aggressive policy on foreclosures or bad loans. This
could result in depressed land prices or in business failures which in turn could weaken other
banks.
This effect has clear policy relevance for bank regulatory policy, although the literature does not
seem to have explored this. Blum and Hellwig (1995)'s model has similar implications for
regulatory policy, but does not operate via the same effect of endogenous asset prices.
4.3 Costs of Financial Distress
A more extreme version of these kinds of effects can occur when banks or non-financial firms
fail. This also provides an instance of the idea that the corporate "veil" may be more substantial.
The idea that financial failures may acerbate economic downturns is, of course, an old one going
back to Fisher (1933) and others. Among the literature reviewed above, it is present in Allen
and Gale (1998) and Bernanke (1983). Together, they suggest that debt may have costs in
aggravating and propagating financial crises (supporting the notion of minimum capital
requirements).
One has to distinguish carefully between financial distress per se and costly financial distress.
Traditionally, financial economics, and particularly the theory of corporate finance, has been
22
careful to maintain this distinction. Modigliani and Miller (1958) emphasized that bankruptcy,
in itself, does not affect the validity of their central result on the irrelevance of capital structure.
What does make debt more costly, however, is when the bankruptcy process is costly and
reduces the value of the firm's business.
Studies for non-financial firms confirm that bankruptcy can be quite costly. For example,
Warner (1977) estimates the legal and other direct fees of a sample of US railroad companies at
4% on average, while other studies suggest the more intangible costs could be several times as
high. See Senbet and Seward (1995) for a survey of research on bankruptcy costs. This work is
based on non-financial firms and there do not seem to be any corresponding empirical studies
for banks. However, one could easily imagine that the costs for a bank could be much higher
than for a typical non-financial firm. Dow and Rossiensky (1997) analyze optimal financial
structure for a bank that faces intangible bankruptcy costs.
23
5. Review of Selected Cases
Having surveyed some of the academic research on systemic risks, I shall now discuss a few
selected cases of failure of financial firms. This is not a systematic overview of cases of failure;
it is based on a selection of cases that seem suggestive. They include some that are archetypal
examples of systemic risk, and others that are not classical cases: they are less well-known and
less frequently discussed. They are also not selected to support any particular theory or model of
systemic risk. However, to give a preview of the discussion later in the paper, one conclusion
that I do draw from these and from other cases, is that the academic literature may have
overemphasized the importance of the propagation mechanism for systemic risk, rather than
source of the systemic risk. This is not to say that the propagation mechanism is unimportant,
but rather that the initial source of systemic risk needs to be studied as well. In fact, my
conclusion is that the problem of moral hazard, broadly defined, is an important source of risk.
These cases are provided as an introduction to the ideas developed later in the paper, where I
discuss models of moral hazard in banking.
5.1 Herstatt
Perhaps the most frequently-cited case of the potential for systemic risk was the Herstatt failure
on 1974. The bank was closed when it became insolvent as a result of taking on large foreign
exchange positions (for references to the Herstatt affair see Financial Times (1974), 27-6-74,
28-6-74, 8-7-74, 9-7-74). The episode took place soon after exchange rates were allowed to
float in the early 1970's, in response to which many banks expanded their Forex trading
operations. 85% of the bank was owned by Hans Gerling and the Gerling Insurance Company;
24
furthermore 5% was owned by the General Partner, Iwan Herstatt, who had unlimited personal
liability for the bank's debts. The failure caused widespread disruption because many Forex
counterparties had made payments to settle recent deals with Herstatt, while Herstatt had not
made the corresponding payments in the other currency. The case is interesting because it
illustrates the dangers of systemic disruption of the payments (in particular, a system with
deferred settlement and without collateral.) However, it also illustrates that an individual trader
may cause huge problems by making trades that are not in the best interests of the bank’s
owners. It is interesting to note that this took place in a firm with a close ownership structure,
even tough economic analysis usually associates such problems with a dispersed ownership
structure that gives inadequate monitoring incentives (e.g. Dewatripont and Tirole, 1994b).
5.2 Swiss real estate crisis of 1991,1992
The Swiss banking industry was affected by a general economic downturn, falling property
markets, and exposure to the bankrupt Omni holding company (in the case of lending to Omni,
several banks exceeded regulations that limit unsecured lending to 20% of own capital). The
main problem was exposure to real estate prices and interest rates: although the majority of
Swiss mortgages are floating rate, banks were hit by rising interest rates as some borrowers
became unable to service their mortgages. The smaller regional and cantonal banks were
particularly hard hit by the economic downturn. A few regional banks were forced to close, e.g.
Spar und Leihkasse Thun, but a rather larger number were taken over. In many of these cases
they were probably insolvent. (For references see Financial Times (1991,92), 18-3-91, 10-4-91,
10-10-91, 16-12-91, 7-2-92, 8-4-92, 5-5-92, 8-8-92.) These events took place at a time when
there was overcapacity in the Swiss banking industry, when traditional practices, and the
25
corresponding rents, had been eroded by increased competition and a more aggressive policy by
the competition authorities. Also, pressures for the large banks to produce more transparent
accounts had led to a demand for higher returns on equity.
The “hidden” exposure to interest rates is highly reminiscent of the effects described by Hellwig
(see section 3.2 above). However, it is interesting that despite the strong negative shock, the
banking industry was able to absorb the impact just as effectively as the rest of the economy. In
other words, the strength of the initial shock to the banking system was large, but it does not
seem to been disproportionately amplified by a propagation mechanism.
5.3 Chilean economic crisis of 1981, 82
The Pinochet regime seized control of four banks (representing almost 20% of all peso deposits)
and four finance companies in late 1981 in a marked reversal of its general policy of laissez-
faire. A large part of the banks' loan portfolio was non-performing and the government cited
"administrative deficiencies" and violations of the banking code as the reason for intervention,
which also included the arrest of two bank directors. The cost of this operation was estimated at
$600bn. Foreign bankers alleged that loan decisions in Chilean banks were often based on poor
criteria, if not outright corruption. Ultimately, in the course of 1982, the financial crisis became
hard to separate from a general economic recession which also saw a shift in the balance of
political power towards the more interventionist forces such as the military (for example in the
course of 1982 there were two changes of central bank governor, as well as several changes in
the composition of the ministerial cabinet). Banks' bad debts by the summer of 1982 amounted
to over $1.5bn, more than half the reserves of the bank sector, and 23 out of 39 banks had either
26
been sold or turned over their portfolio of bad debts to a government-sponsored rescue
operation (for references see Financial Times (1981,82), 4-11-81, 10-11-81, 18-2-82, 26-3-82,
6-5-82, 15-7-82, 3-8-82, 3-9-82, 17-11-82).
The Chilean episode does have a similarity with the case previously described for Switzerland:
it seems to have been the strength of the basic shocks, rather than the effect of a propagation
mechanism, that caused the problems (this is not to deny that there was also a propagation
effects). An additional element appears to have been a kind of failure of individual incentives,
in that the bad debts accumulated by the banks were probably not the result of actions that
benefited their shareholders, but rather resulted from defective organizational structures.
5.4 UBS and SBC (1997)
The merger between UBS and SBC announced in 1997 was initially presented as a merger of
equals, but it gradually emerged that UBS had been weakened by a number of recent loss-
making events and that the merger was actually an SBC takeover (for references see The
Economist 24-1-98 and 31-1-98 and Euromoney March 1988). Derivatives losses at UBS
related to imperfect hedging of positions in convertible debt securities issued by Japanese banks
amounted to about SFr 650m (for comparison, UBS's own capital was over SFr 10bn.) The
bank also lost about SFr 200m as a result of an unexpected change to British tax regulations for
financial traders, and suffered extremely bad publicity in relation to its treatment of Nazi
Holocaust victims.
The problems at UBS cannot be described as systemic, in the sense of being transmitted to other
27
organizations within the financial system. However, one could say that potentially they could
have had a systemic impact if they had been larger or if UBS had had much greater leverage.
They are also important in the sense that UBS/SBC between them comprise a large part of the
Swiss banking industry. The key feature seems to have a failure of managerial culture, with an
inadequate focus on profitability and transparency.
28
6. Moral Hazard
I now turn to a discussion of the role of moral hazard in systemic risk. There are several
motivations for this. First, the existing literature on systemic risk has given relatively little
emphasis to moral hazard, hence it remains a fruitful area to explore. Secondly, it is evident that
moral hazard is an important element of many financial crises. Beyond the arguments made in
section 5, this connection will be investigated more systematically in section 7 below. Thirdly, I
believe that the kinds of moral hazard that arise in financial crises suggest interesting variants of
the standard principal-agency problem usually studied in the economics literature: the existing
models normally study a single agent and a single principal, where the principal designs the
contract with full knowledge of the trading environment. In contrast, experience of financial
crises suggests the presence of incentive problems at a collective level, and where an explicit
contract design process is absent. In this paper, I will not explicitly formulate alternative
models to represent these situations, but I will explore them in a qualitative and preliminary way.
Before starting on the discussion of moral hazard, it is worth clarifying the role of moral hazard
in relation to traditional banking research. Most traditional research has assumed, implicitly or
explicitly, that the financial system is a unique part of the economy because relatively
unimportant shocks are amplified by a powerful propagation mechanism. Hence the source of
the shocks has received less attention that the functioning of the different possible propagation
channels. In contrast, moral hazard is not a propagation mechanism but a mechanism for
initiating shocks. This raises a number of very interesting and interrelated questions: are the
shocks correlated, so that in aggregate they can be large even without a propagation
mechanism? Are these shocks multiplied by leverage and the use of derivatives, with the same
29
effect? Indeed, are banks qualitatively unique or are they simply very highly levered firms? A
full exploration of these questions is beyond the scope of the paper, but they should be borne in
mind throughout the discussion that follows.
Research on banking has emphasized the moral hazard for bank equity-holders to expropriate
debtholders (or the deposit insurer) by asset substitution (i.e. holding a riskier portfolio of
assets). Undoubtedly, this emphasis has been reinforced by the experience of the USA in the
1980's, where many S&L's engaged in risky exposure to the yield curve. However, recent
experience, as well as experience from earlier periods, has pointed to the importance of two
other types of moral hazard:
1) the possibility that individual traders will take actions (specifically, excessive risk-taking) at
the expense of the firm as a whole;
2) the possibility that bank management will not act in the shareholders' interests.
6.1 Incentive Problems for Traders
I shall consider these two kinds of moral hazard in turn, starting with the moral hazard or
agency problem for individual traders or small groups of traders. This case seems the easier of
the two to analyze. In the first place, most traders have a bonus or inventive element to their
remuneration that includes an option-like element: in other words, they share in the profits they
make (above a given threshold), but not in the losses. To maximize their expected remuneration
therefore, they should take on as much risk as possible. Even when this consideration is
30
balanced with an aversion to risk in their remuneration, traders will still have a strong incentive
to take on risky positions. What is more, this incentive will become progressively stronger as
their trading position deteriorates below the level required to earn a bonus (let alone when it
deteriorates below the level at which they would be sacked). Hence a trader who has performed
badly has an incentive to take more and more risky positions ("gambling for resurrection"), as
well, of course, as to manipulate his accounts to hide the losses. One of the best-known
examples of this problem in recent years was Nick Leeson at Barings (Hogan, 1996).
6.2 Why is traders' remuneration convex?
A remuneration package consisting of a base salary plus a constant share of profits above a
given threshold is an example of a convex payoff function (a function f(x) is convex if it always
lies above its tangent, or equivalently f(αx+(1-α)y) ≤ αf(x) + (1-α)f(y)). By Jensen's inequality,
the expectation of a convex function of wealth is increased by adding on a pure, actuarially fair,
gamble (i.e. Ef(x) ≥ f(Ex)). In other words, the expected value of a convex payoff such a bonus
scheme will be increased by adding risky, zero-NPV trades to the trading position. Another way
of putting this is to note that a base salary plus bonus equates to giving the trader a call option
on the value of his position. The Black-Scholes value of an option increases in the variance of
the underlying asset; hence, the value of the bonus option increases if the underlying trading
position becomes riskier.
This can be misleading. Traders care about the risk of their bonus as well as the expected value.
It is true that the expected value of a convex bonus payment increases in the risk of the position,
but the risk of the payment will increase also. Or, to put it in terms of option pricing theory, the
31
Black-Scholes formula gives the valuation of the option to someone who can hedge the risk of
the option by trading in the underlying security. Since options are highly risky, someone who is
risk-averse and cannot trade directly in the underlying security will value an option much less
than the Black-Scholes value, and will not always welcome an increase in risk. In addition,
there is a dynamic aspect to the problem since poorly performing traders will tend to be fired (or
- depending on the culture of the country -transferred to other tasks within the firm). This
dynamic aspect is not captured by viewing remuneration incentives as a single-period option.
Hence, traders' bonuses will not necessarily always induce an excessive appetite for risk-taking -
- though there is unquestionably a danger that they may do so.
This tendency for bonuses to increase the incentive to take risks is well understood. This raises
the question of why traders' compensation is structured this way. Is it optimal to give traders
convex payoffs? Why should it be optimal to encourage them to take on risk in this way? If it
is not optimal, why are they given convex payoffs?
This is an interesting question that has several dimensions. Generally speaking, research has
ignored the question of optimality, and confined itself to noting the risk incentive problems
arising from bonuses. Research into optimal contracts, on the other hand, has generally
predicted remuneration contracts that are not bonus-like (Bhattacharya and Pfleiderer (1985),
Stoughton (1993)). There are two ways in which observed (bonus-like) contracts differ from
optimal contracts in the literature. First, the optimal contracts are not always increasing. A
trader whose profits are "too large" will be penalized: this is because such profits are more likely
to arise by mistake (as the result of a wildly inaccurate prediction of asset values) rather than as
a result of a good analysis of the situation. This is the case in Bhattacharya and Pfleiderer
32
(1985) for example, where the optimal rewards function is quadratic. In reality, however, one
can point to many cases where traders were simply rewarded for very large profits when, in
retrospect, managers would have done better to react negatively and to investigate the source of
these profits more carefully (it is only fair to say that managers of financial forms are well aware
of this problem and invariably claim that they do take it into account).
Secondly, the optimal contracts in the literature involve payments from the trader to the firm (in
the case of bad performance) as well as payments from the firm to the trader (in the case of good
performance). It is not clear why such contracts are rarely seen in practice. Some firms have
tried to force traders' bonuses to be retained within the firm for some years - effectively allowing
part of the bonus to be reclaimed later in case of poor performance - but these efforts have met
with resistance in the labour market.
Dow and Gorton (1997) present a model of optimal contracting where traders cannot be forced
to make payments to the firm in case of poor performance, and hence the optimal contract has a
bonus-like element. In that model, even though the option-like contract is optimal among the
feasible alternatives, it induces traders to trade excessively and take on more risk.
6.3 Incentive Problems for Bank Management
The above discussion of agency and incentive problems has focused on individual traders (or
trading desks). As such it is within the standard incentive-contracting paradigm. However, as I
have argued above, an interesting and somewhat different problem is the problem of incentives
for bank management: does bank management have an incentive to take on too much risk? This
33
problem seems more important because after all, every time an individual trader takes excessive
risk, there is a failure of management that did not prevent, or may have encouraged, the risk-
taking. Post-mortems of notorious trading losses such as Barings 1994, Herstatt 1976, Nat West
Markets 1997, UBS 1997 tend to emphasize that the failure is as much a failure of control on
the part of management, as a failure of a single trader's position.
On the other hand, losses that cannot be ascribed to out-of-control individuals (or small groups)
are often blamed on poor management. For example, Yamaichi's 1997 failure was the
culmination of a series of structural problems in the firm's corporate culture. Among these,
exposure of the firm's connection with sokaiya racketeers dealt a humiliating blow to its prestige
and to its management's credibility. Perhaps equally important was the exposure of accounting
irregularities resulting from Yamaichi acceding to compensation demands from corporate
clients for trading losses.
Among the examples I gave in section 5 above, the problems in Switzerland in 1991-92 and in
Chile in 1981-82 seem less specific and it is harder to point to a source of management failure.
However, they both share the feature that the corporate culture apparently produced a system
with some inertia that was not responsive to change.
This aspect of moral hazard has not been addressed much in research, yet seems an important
priority. Jensen (1986) has promoted the idea of "Free-cash-flow theory," namely that firm
managers may be motivated to increase the size of the firm even when this is not in the interests
of equityholders. This is similar to the idea of a weakness in corporate culture, but still treats
the bank management as if it were a single decision-making agent. In reality, the problem of a
34
weak or defective corporate culture seems to have inherent collective characteristics. Within a
group of managers, there may be inadequate incentives for an individual to diagnose and draw
attention to problems, and to implement solutions to those problems. This inertia can operate at
the level of the firm, or indeed at an industry-wide level.
6.4 Are managers rational expected-utility maximizers?
Economic theory has a well-developed set of tools to analyze (individual) incentive problems
using expected-utility theory. As in other areas of economics, insight and predictions are based
on rational, utility maximizing behaviour. However, it is worth pointing out that this
assumption may not always be warranted. Particularly in relation to behaviour in the face of
risk, standard economic behavioural assumptions may be inaccurate. People may face distorted
perceptions of risk and they may make distorted evaluations of the consequences of risky
decisions.
There has traditionally been relatively little research on irrationality in financial markets, but it
is now beginning to be accepted and recognized as a rich and important area for investigation.
For example, Daniel, Hirshleifer, and Subrahmanyam (1998) investigate the theoretical
implications of investor irrationality for asset pricing.
Fenton O’Creevy, Nicholson,Willman, and Soane (1999) investigate the behaviour of financial
market traders in simulated trading environments, using a sample of traders from various firms
in the City of London. They find that traders display systematic overconfidence in their ability
to control their portfolio returns, a phenomenon recognized and studied by psychologists and
35
known as “illusion of control.” Furthermore, traders with higher illusion of control have
systematically lower trading profits.
In addition, Economics traditionally defines rational behaviour in a rather narrow way. Hence
principal-agent models focus exclusively on the monetary dimension of incentives. Nicholson,
Willman, O’Creevy and Soane (1999) discuss how incentives for traders depend on other social
and psychological influences, and explain how these influences can cause excessive risk-taking.
36
7. Which kind of systemic risk is most important?
As I have argued above, "systemic risk" is often used as a catch-all phrase to cover many
different kinds of problem. The academic literature has developed several research tropes that
bear on this question. However, when one looks at particular episodes of financial failure, these
do not necessarily seem to be the most relevant ones. For example, the Herstatt collapse posed
grave problems of settlement risk but was also a problem of unauthorised trading, i.e. a moral
hazard problem. The recurrent problem of excessive risk-taking in bank loan portfolios is also
often a moral hazard problem, e.g. in the S&L crisis of the US in the 1980's or in the Credit
Lyonnais debacle - in the latter case, the moral hazard taking a broader, more political
dimension in that neither the bank management nor the French state appears to have been
concerned to implement a more disciplined approach to making loans. Looking at these and
other cases, it would seem that moral hazard, broadly defined, is the single unifying thread that
seems to run through them. Here I will try to make precise which kinds of moral hazard seem
most relevant by surveying a larger number of cases (they include various problem cases as well
as actual failures). These are listed in table 1. (Although the table was compiled from press
cuttings, Steinherr (1998) is an excellent source for descriptions of many of these events.)
37
Table 1
Episode Nature of Problem
Sumitomo (1996) RTMetallgesellschaft (1993) RTBarings (1994) RTYamaichi (1997) PCCYamaichi (1965) PCCSweden (1990) COEHerstatt (1974) RTSwitzerland (1991-2) COE (PCC?)Continental Illinois (1984) ?Chile (1981-2) COE, PCCCredit Lyonnais (1990's) PCCNatwestmarkets (1997) RTUSA S&L crisis (1980's) ACCGibson Greetings - Bankers Trust (1994) ACCProcter and Gamble - Bankers Trust (1994) ACCSalomons (1994) ACCUBS (1997) PCCDrexel Burnham Lambert (1990) ACCCounty Natwest (1997) ACC
Key:
RT: "rogue trader:" excessive risk taking or fraudulent behaviour by trader or small group oftraders, together with a failure of management control that fails to detect them.
PCC: "passive corporate culture:" management culture of the firm is inadequately focused onsound profitability and transparency.
ACC: "aggressive corporate culture," the firm's focus on profitability is so strong that traderstake excessive risks to enhance short-term profitability.
COE: "collective overexposure" a large number of banks are, collectively, heavily exposed tothe same risk (even though each individual bank's position may not seem particularly imprudentex-ante).
Note: Most instances of RT, ACC, and PCC are cases of the general moral hazard problem.
38
After considering the cases in the table, I have tried to categorize the nature of the problem. The
categories chosen have been selected after looking down at all the cases. It seems that in many
cases, some form of inadequacy of incentives was at work. However, one cannot say that they
are all examples of a similar kind of moral hazard problem. Rather, the deficiencies of the
incentive system take different forms that I have tentatively classed into three groups. The first,
and simplest, kind of moral hazard problem is a simple "rogue trader" problem. These are all
cases where a single trader (or a small group) had a disproportionate amount of discretion and
was able to use this power to build up large losses and conceal them. Of course, each such
instance also implies a failure of control at a higher level of management, but in most of these
cases one cannot generally identify this management failure except with the benefit of hindsight.
In other words it is not easy to point to a general distortion of incentives within the firm which
led to the problem, rather it is a case of being able (ex-post) to document some specific, often
serious, control lapses which permitted the trader to continue accumulating large losses. I have
denoted these cases RT in the table. They include the Sumitomo copper trading losses,
Metallgesellschaft, Barings, Herstatt, etc.
In some other cases, one cannot point to such a clear failure of control, nor to a single individual
as the main instigator of the problem. It would seem in these cases that the firms had an
aggressively profit-oriented culture which encouraged, or at least permitted, some employees to
overstep the mark. The culture facilitated or sometimes encouraged excessive risk-taking. I
have denoted these cases ACC, "Aggressive Corporate Culture," for want of a better description.
This group is actually rather varied: for example I have included Drexel Burnham Lambert
which is intimately identified with a single individual. I have also included the US S&L crisis
of the 1980's where the stimulus to take risks came primarily from the commercial and
39
regulatory environment, i.e. from outside the firms themselves.
Although the second group is rather heterogeneous, it is interesting chiefly to provide a contrast
with a third group. This third group is what I call "passive corporate culture," (PCC). Recall
that the first group of cases (RT) corresponded fairly closely to the standard agency and moral
hazard problems that have been studied in the academic literature, i.e. a single agent whose
incentives, in case of failure of the firm's control mechanisms, may encourage risky or
suboptimal actions. Generally, the problem has been that the trader's incentives were in some
sense "too strong," in other words, the trader's incentives (such as convexity arising from the
bonus) led him to try hard to make high profits but without giving enough weight to the
consequences of failure. The other two groups (ACC and PCC) involve a kind of collective
moral hazard encompassing a group of people rather than a single individual. This situation has
not been studied intensively in the academic literature, unlike the single-agent moral hazard
problem. What is interesting is that many of the instances of this collective moral hazard
problem do not involve excessively strong incentives to make profits. Some of them involve
excessive inertia in companies where lack of transparency co-exists with long-term low
profitability.
Finally, one can also distinguish a fourth category that I have called "collective over-exposure."
These cases seem to capture some of the features of Hellwig's analysis described in section 2
above: namely, the problem was not that an individual bank was overexposed to a particular risk.
Rather, it was that the whole system turned out to be exposed and hence no individual bank had
a strong incentive to correct the problem.
40
8. Conclusions
In this paper I have approached the economics of systemic risk in two ways. First, I discussed
the existing literature that is related to the problem of systemic risk. Secondly, I then turned to a
study of a number of cases where problems arose with the financial sector and with individual
firms. This led to the discussion of the role of moral hazard in financial crises.
Research has emphasized the simultaneous determination of asset prices and the possible
macroeconomic consequences of this. One key assumption in much of this literature is that
banks are capital constrained, or strongly prefer debt financing to equity. Although this may be
in keeping with facts about how banks do finance themselves, it is also a worrying assumption
because we do not have a full economic understanding of why this should true.
A review of selected cases suggests that the vast majority of cases are caused by moral hazard,
broadly defined. However, problems caused by a single trader with distorted incentives are in
the minority. A more common problem is a collective failure of management in the bank (or
across the financial system), leading to inertia and the inability to respond to changed economic
circumstances.
This suggests several interesting conclusions as well as pointing to areas for future research.
One conclusion is on the relative role of the mechanism that initiates shocks to the financial
system and the mechanism that propagates them. Traditional banking research is based on the
premise that banking is fundamentally different from other industries and that the financial
system contains a powerful propagation mechanism that can amplify small initial shocks.
41
Implicitly, the moral hazard approach suggested here is based on the idea that initial shocks are
amplified within the firm by moral hazard and by financial leverage, so that the shock delivered
to the financial system by the individual firm can itself be large. Hence systemic risk does not
necessarily depend a powerful propagation channel that multiplies up the effects of small shocks.
The propagation effects may or nor be large, but the initial shock itself can large enough to
cause systemic harm. The LTCM crisis is an example.
This perspective also seems interesting because it suggests that banks may not be,
fundamentally, economically unique institutions– or rather, what makes them unique is simply
their ability to use leverage and to multiply their risk exposure.
The other idea that seems promising for future research is the formulation of models of what I
have called “collective moral hazard” – while the precise definition of what I mean by this may
be not possible before actually specifying such a model, what I have in mind is a situation where
many agents are involved and no single agent has the role of designing a contract as a standard
constrained optimization problem.
42
9. References
Allen, F. and D. Gale (1998), "Bubbles and Crises," working paper, Wharton School.
Bagehot (1873) W., "Lombard Street," King, London.
Bernanke, B. (1983), "Non-monetary Effects in the Propagation of the Great Depression,"
American Economic Review 73:257-276.
Bernanke, B. and M. Gertler (1989), "Agency Costs, Net Worth, and Business Fluctuations,
American Economic Review 79:14-31.
Bernanke, B., M. Gertler and S. Gilchrist (1996), "The Financial Accelerator and the Flight to
Quality," Review of Economics and Statistics 78:1-15.
Bhattacharya, S. and P. Pfleiderer (1985), “Delegated Portfolio Management,” Journal of
Economic Theory 36: 1-25.
Blum, J. and M. Hellwig (1995), "The macroeconomic implications of capital adequacy
requirements for banks" European Economic Review 39: 739-749.
Bolton, P., and X. Freixas, (1994), "Direct Bond Financing, Financial Intermediation and
Investment," working paper.
Boot, A. and A. Thakor (1997), "Financial System Architecture," Review of Financial Studies,
Cox, J., J. Ingersoll and S. Ross (1985), "A Theory of the Term Structure of Interest Rates,"
Econometrica, 53, 385-407.
Daniel, K., Hirshleifer, D. and Subrahmanyam, A. (1998), “Investor Psychology and Security
Market Under- and Overreactions,” Journal of Finance 53: 1839-85.
43
Dewatripont, M. and J. Tirole (1994a), "A Theory of Debt and Equity: Diversity of Securities
and Manager-Shareholder Congruence," Quarterly Journal of Economics 109, 1027-54.
Dewatripont, M. and J. Tirole (1994b), The Prudential Regulation of Banks, MIT Press.
Diamond, D. and P. Dybvig (1983), "Bank Runs, Deposit Insurance, and Liability," Journal of
Political Economy, 91, 401-419.
Dow, J. and G. Gorton (1997a), "Stock Market Efficiency and Economic Efficiency: is there a
connection?," Journal of Finance, 52, 1087-1129.
Dow, J. and G. Gorton (1997), "Noise Trade, Delegated Portfolio Management, and Economic
Welfare," Journal of Political Economy, 105, 1024-1050.
Dow, J. and N. Rossiensky (1997), "The Cost of Capital for Financial Firms," working paper,
European University Institute.
Fenton O’Creevy, M. (1999) “Trading on Illusions: unrealistic perceptions of control and
trading performance,” working paper, European University Institute.
Fisher, I. (1933) "The debt-deflation theory of great depressions," Econometrica 1: 337-357.
Financial Times (1974) "Herstatt Bank Closed," 27 June 1974.
Financial Times (1974) "German Markets Hit by Collapse of Herstatt," 28 June 1974.
Financial Times (1974) "Prosecutor to Start Proceedings for Possible Company Law
Infringements," 28 June 1974.
Financial Times (1974) "BIS to discuss health of Financial Markets," 8 July 1974.
Financial Times (1974) "Central Banks now Ready to Act as Last-resort Lenders," 9 July 1974.
44
Friedman, M and A. Schwartz (1963), A Monetary History of the United States, 1867-1960,
Princeton University Press.
Freixas, X. and B. Parigi (1997), "Contagion and Efficiency in Gross and Net Interbank
Payment Systems.
Freixas, X. and J.-C. Rochet (1987), "The Microeconomics of Banking," MIT Press.
Gertler, M. and S. Gilchrist (1993), "The Role of Credit Market Imperfections in the Monetary
Transmission Mechanism, Scandinavian Journal of Economics 95: 43-64.
Gertler, M. and S. Gilchrist (1994), "Monetary Policy, Business Cycles, and the behaviour of
Small Manufacturing Firms," Quarterly Journal of Economics 109: 309-340.
Greenwald, B. and J. E. Stiglitz (1990), "Macroeconomic Models with Equity and Credit
Rationing," in Hubbard (1990).
Hellwig, M. (1991), "Banking, Financial Intermediation and Corporate Finance," in A.
Giovannini and C. Mayer (eds.), European Financial Integration, Cambridge University
Press.
Hellwig, M. (1994a), "Liquidity Provision, Banking and the Allocation of Interest Rate Risk,"
European Economic Review 39: 1363-1389.
Hellwig, M. (1994b), "Banking and Finance at the End of the Twentieth Century," working
paper, WWZ, University of Basel.
Hellwig, M. (1995), "Systemic Aspects of Risk Management in Banking and Finance," Swiss
Journal of Economics and Statistics, 131, 723-737.
Hellwig, Martin (1998), "Banks, Markets and the Allocation of Risk in an Economy" by Martin
Hellwig, Journal of Institutional and Theoretical Economics (Zeitschrift für die gesamte
45
Staatswissenschaft), forthcoming March 1998.
Hogan, W. P. (1996), "The Barings Collapse: Explanations and Implications," working paper
235, University of Sydney.
Holmstrom, B. and J Tirole (1997), "Financial Intermediation, Loanable Funds and the real
sector," Quarterly Journal of Economics, 112, 663-691.
Holthausen, C. (1997), "Systemic Risk, Interbank Lending and Monetary Policy: a Literature
Review," working paper, University of Freiburg.
Hubbard, R. G. (1990), ed., Asymmetric Information, Corporate Finance and Investment,
University of Chicago Press.
Jensen, M. C. (1986), "Agency Costs of Free Cash Flow, Corporate Finance and Takeovers,"
American Economic Review, 76(2), 659-665.
Kashyap, A. K., J. C. Stein and D. W. Wilcox (1993), "Monetary Policy and Credit Conditions:
Evidence from the Composition of External Finance," American Economic Review 83:
78-98.
King, M. (1994), "Debt Deflation: theory and evidence," European Economic Review 38: 419-
445.
Kiyotaki, N. (1997), "Credit and Business Cycles," working paper, London School of
Economics.
Kiyotaki, N. and J. H. Moore (1997a), "Credit Cycles," Journal of Political Economy, 105: 211-
248.
Kiyotaki, N. and J. H. Moore (1997b), Credit Chains, working paper, London School of
Economics.
46
Marsh, P. (1979), "Equity Rights Issues and the Efficiency of the UK Stock Market," Journal of
Finance, 34, 839-862.
Matten, C. (1996), Managing Bank Capital: Capital Allocation and Performance Measurement,
John Wiley & Sons, Chichester, England.
Mayer, C. (1998), "Financial Systems and Corporate Governance: a Review of the International
Evidence," Journal of Institutional and Theoretical Economics, forthcoming March
1988.
Mayer, C. (1990) "Financial Systems, Corporate Finance and Economic Development," in
Hubbard (1990).
Miller, Merton H. (1995), "Do the M&M propositions apply to banks?" Journal of Banking and
Finance, 19: 483-449.
Modigliani, F. and M. H. Miller (1958), "The Cost of Capital, Corporation Finance and the
Theory of Investment," American Economic Review, 48, 261-297.
Myers, S. C. and N. S. Majluf (1984), "Corporate Financing and Investment Decisions when
Firms Have Information Investors do not Have," Journal of Financial Economics, 13,
187-221.
Nicholson, N., P. Willman, E. Soane and M. Fenton O’Creevy, “The performance of traders and
the competitiveness of investment banks,” working paper, Centre for Organizational
Behaviour, London Business School.
Oliner, G. and Rudebusch G. (1993), "Is there a Bank Credit Channel to Monetary Policy,"
working paper, Federal Reserve Board of Governors.
Raposo, C. (1999), “Essays on the Theory of Corporate Risk Management,” PhD thesis,
London Business School.
47
Rochet, J.-C. (1992) "Capital requirements and the behaviour of commercial banks," European
Economic Review 36: 1137-1178.
Rochet, J.-C. and J. Tirole (1996) "Controlling Risk in Payments Systems," Journal of Money,
Credit and Banking 28: 832-862.
Rochet, J.-C. and J. Tirole (1996) "Interbank Lending and Systemic Risk," Journal of Money,
Credit and Banking 28: 733-762.
Schaefer, Steven M. (1990), "The Regulation of Banks and Securities Firms," European
Economic Review 34: 587-597.
Senbet, L. W. and J. K. Seward (1995), “Financial Distress, Bankruptcy and Reorganization,” in
R. A. Jarrow, V. Maksimovic and W. T. Ziemba, eds. Finance (Handbooks in
Operations Research and Management Science vol. 9), Elsevier Science, Amsterdam.
Steinherr, A. (1998), Derivatives: the wild beast of finance, Wiley, Chichester.
Stiglitz, J. and A. Weiss (1981), "Credit Rationing in Markets with Imperfect Information,"
American Economic Review 71: 393-410.
Stoughton, N. (1993), "Moral Hazard and the Portfolio Management Problem," Journal of
Finance 48: 2009-2028.
Suarez, J. and O. Sussman (1997a), "Endogenous Cycles in a Stiglitz-Weiss Economy," CEPR
discussion paper 1604, forthcoming Journal of Economic Theory.
Suarez, J. and O. Sussman (1997b), "A Stylized Model of Financially-Driven Business Cycles,"
working paper, CEMFI and Ben-Gurion University.
Summers, B. J. (1996), "Comment on Controlling Risk in Payments Systems," Journal of
Monetary Economics 28: 863-869.
48
Suzuki, Yoshio (ed.) (1987), "The Japanese Financial System," Oxford University Press.
Swiss Journal of Economics and Statistics (1995), "Capital Adequacy Rules as Instruments for
the Regulation of Banks," special issue, proceedings of a conference held at Basel on
June 7 1995, vol 131 no 4/2.
Tirole, J. (1994) "On Banking and Intermediation, European Economic Review 38: 469-487.